Health-Related Social Disengagement in Elderly Diabetic Patients: Association with subsequent disability and survival

Data are from the Medicare Health Outcomes Survey (HOS), a longitudinal cohort mail and telephone survey of Medicare beneficiaries enrolled in Medicare+Choice (managed care) plans in the U.S. Medicare HOS was sponsored by Health Care Finance Administration (currently the Center for Medicare and Medicaid Services [CMS]) and developed by National Committee for Quality Assurance (NCQA), a body responsible for accreditation of managed care plans. The Medicare HOS was designed to better evaluate care quality by incorporating key health outcome measures outlined in the Health Plan Employer Data and Information Set (HEDIS) in 1998.

Annual participation of Medicare Managed Care Organizations in the HOS has been mandated by CMS since 1998. All Medicare managed plan enrollees, who have been continuously enrolled for at least 6 months, were eligible for sampling, except those who were on Medicare as a result of end-stage renal disease. For baseline survey, in managed care plans with more than 1,000 enrollees, the survey instruments were given to a random sample of 1,000 subjects per plan; in plans with 1,000 enrollees or fewer, the surveys were conducted for all eligible subjects. A full description of the Medicare HOS, rationale, methods, survey design, sample selection, and respondents’ characteristics can be found elsewhere (28–31).

For the 2-year follow-up survey in 2000, subjects were restricted to enrollees participating at baseline survey who were still alive and enrolled in the original managed care plan. We used an analytic data file from the Medicare HOS Cohort 1 Baseline (1998) and the Cohort 1 Follow-Up survey 2 years later. Because many managed care organizations discontinued or consolidated their health plans between 1998 and 2000, the analytic data at the 2-year follow-up comprised 188 managed care markets and 134,076 completed surveys (30).

The NCQA-accredited HOS interviewers used standardized questionnaires, survey letters, and follow-up telephone calls (as outlined in the HEDIS protocol, including a Spanish language version) to collect information on several variables including physical and mental health functioning using SF-36 and ADL impairments using specific symptoms related to various organs, presence of chronic medical conditions, measure of pain, cancer treatment, smoking, and self-rated health and sociodemographic variables.

This study reported on diabetic subjects aged ≥65 years with no ADL disability at baseline (n = 8,949). After excluding 155 subjects with incomplete data on baseline social functioning, 8,794 subjects were analyzed for the outcome of death. Due to missing data on social functioning (n = 53) and ADL (n = 3), our final sample (n = 5,232) for the ADL analysis comprised those subjects who completed the 2-year follow-up interviews and who had complete data on baseline social functioning and follow-up ADL. Causes of nonparticipation at the 2-year follow-up surveys were disenrollment (n = 2,443), death (n = 471), and refusal or inability to contact (n = 700). Forty-seven subjects with no available data were coded in the Analytic Database as “invalid.” The characteristics between the subjects who completed the 2-year follow-up and the subjects who dropped out from follow-up were compared. There was no significant difference in social functioning scores between the two groups. However, the higher dropout rates were found among nonwhites, subjects with depression, and subjects with cancers and chronic obstructive pulmonary disease (COPD).

Diagnosis of diabetes and other comorbidities were by self-report. We used information on relevant sociodemographic variables, chronic medical conditions, social functioning component of SF-36 and other variables at baseline, and data on ADL and mortality at the 2-year follow-up interview (2000), as contained in the Cohort 1 Analytic Public Use File published by CMS, Health Services Advisory Group, in January 2003 (30).

The SF-36 is a 36-item measurement instrument used to assess health status and health-related quality of life (32,33). The SF-36 survey instrument (the whole instrument or its subscales) is among the most frequently used health status measures in studies of older adults (31–35). It has two summary scales, the Physical Component Summary and the Mental Component Summary, which were derived from eight subscales: physical functioning, role-physical, bodily pain, general health, vitality, social functioning, role-emotional, and mental health.

The social functioning subscale (range 0–100) was derived from responses to two questions such as, “During the past 4 weeks, to what extent has your physical health or emotional problems interfered with your normal social activities?” (responses range from not at all to slightly, moderately, quite a bit, and extremely) and “During the past 4 weeks, how much of the time has your physical health or emotional problems interfered with your social activities?” (responses range from all, most, some, a little, to none of the time). Lower scores on any or all of the scales indicate poorer health functioning. Standardized scores on SF-36 for the U.S. population are used in most studies to calculate a normalized score. A score of 50 represents the national average for the summary scores and for the subscales. A score that is 10 points above or below the mean score of 50 represents a difference of 1 SD from the national average.

There were six ADL questions in the Medicare HOS. These included walking, bathing, dressing, eating, getting in or out of chairs, and using the toilet. Respondents were asked to indicate whether they could perform these activities without help, if they needed help, or if they were unable to do them. “No disability” was defined as needing no help, and “any disability” was defined as needing help with or unable to perform one or more of the six ADL.

The indicator of decease (without the specific date of death) at the 2-year follow-up interview (2000) was from the Cohort 1 Analytic Public Use File published by CMS, Health Services Advisory Group, in January 2003 (30).

Factors potentially associated with disability and mortality among individuals with diabetes include sociodemographic variables (age, sex, ethnicity, marital status, years of formal education, smoking status), comorbidities (hypertension, angina or coronary artery disease, myocardial infarction, stroke, COPD, arthritis, cancer), and history of depressive symptomatology, assessed by a positive screen on a depression questionnaire (36). Positive depression screen was assessed by an affirmative answer to any of the following questions: 1) “In the past year, have you had 2 weeks or more during which you felt sad, blue, or depressed, or when you lost interest or pleasure in things that you usually cared about or enjoyed?”; 2) “In the past year, have you felt depressed or sad much of the time?”; 3) “Have you ever had 2 years or more in your life when you felt depressed or sad most days, even if you felt okay sometimes?”

We examined demographic and health factors for diabetic subjects, stratified by ADL disability and death at 2-year follow-up, using descriptive and t tests for continuous variables and contingency tables (χ²) for categorical variables. Two logistic regression models were used to predict the risk of ADL disability and death. The first model is unadjusted; the second model added covariates that were significantly associated with ADL limitation or death in the bivariate analyses, i.e., age, sex, marital status, smoking status, depression, comorbidities, and general health from SF-36. All analyses were estimated using the SAS System for Windows, version 8.2 (SAS Institute, Cary, NC).

Descriptive demographic and health factors for initially nondisabled older diabetic subjects, stratified by incident ADL disability and death at 2-year follow-up, are shown in Table 1. Significant bivariate correlates of ADL disability and death among older diabetic subjects included older age, being unmarried, depression, history of myocardial infarction and other heart condition, stroke, COPD, arthritis of hip or knee, cancer, and increasing number of comorbidities. Ethnicity was not significantly associated with ADL disability or mortality.

The difference in baseline social functioning scores according to ADL disability and death were also analyzed (data not shown). Baseline social functioning scores were significantly lower among the disabled and the deceased (mean ± SD, 49.9 ± 9.5 vs. 52.9 ± 7.5 for disabled and nondisabled, P < 0.001, and 49.0 ± 10.5 vs. 51.8 ± 8.5 for deceased and alive, respectively, P < 0.001).

Table 2 shows, in the unadjusted model (model 1), a significant association between higher scores on social functioning measures (as a continuous variable) and increased risk of ADL disability at the 2-year follow-up (P < 0.001). After full adjustments were made for age, sex, marital status, smoking, myocardial infarction and other heart conditions, stroke, COPD, arthritis, cancer, depression, and general health of the SF-36 (model 2), increasing social functioning scores were still predictive of lower odds of ADL disability at the 2-year follow-up. For each 10-point increase in social functioning score at baseline, older diabetic subjects in our study experienced 18% lower risk of any ADL disability (odds ratio [OR] 0.82, 95% CI 0.75–0.89; P < 0.001).

The results of two logistic regression analyses predicting the risk of death are also shown in Table 2. In model 1, increasing scores on social functioning measures at baseline were significantly associated with decreased odds of death (P < 0.001). With full adjustment for age, sex, marital status, education, depression, comorbidities, and general health (model 2), better social functioning at baseline was associated with decreased risk of 2-year mortality. For each 10-point increase in social functioning score at baseline, older diabetic subjects in our study experienced 12% lower risk of death (OR 0.88, 95% CI 0.78–1.00; P = 0.043) over a 2-year period.

Our findings can be summarized as follows: among community-dwelling initially nondisabled diabetic subjects aged ≥65 years in the Medicare Managed Care program, higher scores for measures of social functioning (less health-related social disengagement) at baseline were significantly associated with better survival and less ADL disability over a period of 2 years, controlling for age, sex, marital status, depression, general health, and comorbidities. There was a stepwise decrease in the risk of future ADL disability and death with better scores on social functioning at baseline, such that with each 10-point increase in social functioning score (less health-related social disengagement), older diabetic subjects in our study experienced 18% lower risk of any ADL disability and 12% lower risk of death over 2 years of follow-up.

Prior research has reported on associations between lower level of social activities and increased risk of disability and mortality (9,12,14–16,18,19,37). For example, data from the New Haven Established Populations for Epidemiologic Study of the Elderly showed that increased participation in social activities (such as visits to theaters, game playing, shopping, and church attendance, among others) was associated with lower risk of physical disability (19). Similarly, Lennartsson et al. (16), using data from the Swedish Panel Study of Living Conditions of the Oldest Old, showed that increased participation in social activities was associated with reduced risk of death at 4-year follow-up, adjusting for health variables and other relevant confounders. Our results extend these studies to older initially nondisabled individuals with diabetes across the U.S. by showing that health-related social disengagements are associated with higher risk of subsequent disability and death, adjusting for relevant confounders. In other words, among elderly diabetic subjects with no self-reported ADL disability, a decrease in social activities might be a forerunner of subsequent loss of independent living and premature death.

The mechanism for the association of participation in social activities with lower disability and mortality is not clear. Past studies suggested that engagement in social activities might be associated with several factors, including an optimal sense of self-competence and self-efficacy, slower cognitive decline, and an improved blood markers of immune and stress response, factors known to improve the odds for disability-free survival (38–41). Our study included no data on measures of self-efficacy, blood markers, and other factors to explore these factors as possible mediators for the link between health-related social disengagement and our outcome measures.

On the other hand, it is possible that decreasing participation in social activities by nondisabled individuals with diabetes is a marker for severity of comorbidity, not just mere presence of a disease. Therefore, an unrecognized medical condition (i.e., subclinical stroke or subclinical depression) might be causing a decrease in both the ability (and the motivation) to engage in social activities and the ability to perform ADL. In that scenario, the decrease in social activities is captured much earlier by assessing patients’ responses to social functioning questions during a period when the subjects reported no ADL disability. However, over a 2-year follow-up period, with progressive decrease in level of social activities (with potential loss of the known health benefits of participation in social activities) as well as the persistence of the unrecognized condition, deficits in ADL performance accumulate, leading to physical disability, loss of independence, and death. We have no direct evidence in our study that incident ADL disability (and death) at follow-up is due to stroke or any specific medical condition that can potentially lead to decrease in participation in social activities and ADL ability. Due to lack of data in our study on severity of diabetes (such as HbA_1c level), we are not able to determine whether health-related interference with social activities is a marker for poor diabetes control among older diabetic patients with no self-reported ADL disability.

Our findings seem to indicate that even when no ADL disability is clinically evident, patients start, perhaps at an early presymptomatic stage of disablement, to adjust their levels of social activities in response to their perceptions of physical and emotional disturbances. Therefore, inquiry about status of social activities in older diabetic subjects by clinicians is one way to identify individuals who will benefit most from interventions to stem the future loss of independence.

This study has some limitations. First, the diagnosis of diabetes and comorbidities were by self-report. However, prior research showed good agreement between self-reported diabetes and diabetes diagnosed by blood tests (42,43). Nonetheless, it is conceivable that subjects might under-report comorbidities (such as peripheral artery disease and peripheral neuropathy), partly because of underdiagnosis and misattribution to aging. It is also possible that, due to diabetes-related nerve damage, the prevalence of self-reported angina could be underestimated. Second, we did not include data on adherence to medication and duration of diabetes, factors known to impact diabetes-related morbidity and mortality (1–4). Medicare managed plan enrollees receive coverage for drugs, and this improved access should minimize the impact of differences in medication availability on our outcomes. Third, the Medicare HOS database has no detailed information on types and frequency of social activities. Fourth, the results of the association between health-related social disengagement and ADL disability might only be generalized to the diabetes population without severe chronic comorbidities of cancer and COPD, among others.

Another limitation of our study is our use of unweighted analyses, which do not account for potential intraclass correlation within health plans. However, to explore potential effects of selection and nonresponse, Hwang et al. (44), using a weighted scheme to adjust for selection and nonrespondent bias in Medicare HOS data, found similar estimates of physical and mental functions on SF-36 between the weighted and unweighted analyses. Nonetheless, the standard errors of estimates were higher in the weighted analyses. Because the dummy identifier of health plans and the weighting schemes were not available in the Cohort 1 Analytic Public Use File, we were not able to perform weighted analysis. Despite these limitations, our source of data has several strengths, including its large sample and its nationwide spread of older enrollees in managed care plans across the U.S.

In conclusion, our study found that health-related limitations of social functioning are significantly predictive of ADL disability and death over a period of 2 years among initially nondisabled older diabetic Americans enrolled in Medicare managed care plans. Multiple approaches are needed to maintain independent living and better survival in elderly diabetic individuals. Older diabetic subjects who report curtailing their social activities might be at high risk of subsequent disability and, as such, should be targeted for more rigorous evaluation and management of any modifiable underlying emotional or physical disorders. Improving disability-free survival in older patients with diabetes requires a better understanding of social functioning and social support in this population. Given the fact that the population aged ≥65 years is one of the fastest growing segments in the U.S., the current findings have important implications for health care planning and resource allocation to maintain independence in older adults living with diabetes.

This study was supported by Agency for Health Research and Quality Grant HS-11618. M.A.R. is supported by the Bureau of Health Professions’ Geriatric Academic Career Award 1K01 HP-00034-01.